It's an alignment AND blending error. It looks like the exposure wasn't kept the same and perhaps not enough overlap. It's best to meter the scene and set your camera to a manual exposure setting that won't blow out highlights in any of the image shots. Overlapping 1/3 up to 1/2 from frame to frame helps prevent alignment errors. I tried a couple of panos with LR Merge to Panorama and got very similar results as Edit In> Merge to Panorama in PS with no alignment errors.
Merge to HDR doesn't use the LR Develop settings prior to merge so you can't correct exposure differences that way. Try adjusting the images using the LR Basic panel Exposure control so they look uniform and then export to 16-bit ProPhoto RGB TIFF. Try creating the panorama using these corrected TIFF files.
Thanks for responding, trshaner, You're quite right there are alignment issues too in this instance. However I don't really think this is down to issues of overlap or exposure in the original shots: all exposures were identical, and overlap more than adequate based on past experience of using Photomerge in PS going back to it first appearance (in PSE2 iirc, then in PSCS)
This screen shot shows exposure exif and also overlap in Library Grid view:
I've also had several successful successful runs with the LR Merge to Pano, but today i've had a couple that failed, and I think there may be a bug lurking in there somewhere. I'll try what you suggest - rendering the files as TIFFs and reimporting to LR then running them through the merge, and report back here tomorrow, but it's not a long-term fix, as it's easier just to go back to the tried & tested Edit in> Open as Layers in PSCC 2014
Thanks for the suggestion, though, it's always good to experiment with these things! But I guess LR6/CC is in de facto public beta, and it may take until 6.1 to get these things fixed
Sometimes matching the exposure is not enough. This may be a problem with not having the white balance the same across all the exposures. I have found this to be a problem where the scene changes from sun to shade. Your photo are not extreme in this sense but maybe enough to throw off the merge. I almost always use manual white balance to eliminate/reduce these types of problems.
Looking at the above screenshot the overlap and alignment is very good, but the right-side two images have Exposure of 1/750, F4.8 and 1/750, F6.7, which is a 1 EV difference. The stitching and blending error is also where these two images overlap. Try my suggestion of Export to full-size ProPhoto RGB 16 bit TIFF and Bob's suggestion of using the same WB for all images. Use your LR defaults (0 Tone settings, Daylight WB) and Sync the four inanges to be sure they are the same. I'm think you'll get better stitching and blending results. You can delete the TIFFs when done.
Hi Bob and trshaner
You raise two points, re WB and Exposure:
WB: as you state Bob, the differences in WB between these shots is pretty minimal : from L to R the WB Temp and Tint values are 5,100 / +18; 5,050 / +19; 5150 / +20; 5250 / +20.
I would suggest the biggest disparity here is between the settings in shots 1 & 2, but it is merging shots 2 & 3 that ACR/ LR is getting it wrong. I also tried just merging the first 3 shots, and just shots 2 & 3, but got the same anomaly. Here is its attempt to merge just shots 2 & 3:
I tried setting equal WB on all pics in the Develop module, but that made no difference. I always use Auto WB in the field unless extreme circumstances suggest I need to use my Expodisc, and I've never had problems of this type when merging pans in recent versions of PS, even when I've been caught out without the Expodisc and had to use Auto WB in less-than-ideal circumstances
Exposure: wrt aperture values, the Leica M series digital cameras have no electronic connection to their lenses, and so have no way of knowing the aperture setting of any lens at any time. So aperture settings the exif data from Leica M's are guesstimates by the cameras' onboard computers. I set the aperture for these shots (probably between f/5.6 and f/8), and did not change it between shots (something that's almost impossible to do by accident on these cameras while shooting handheld)
As the preview appears in the Merge to Pano window, there is a point at which that "crack" suddenly appears, just a 1/2second before it says "job done". So I'm convinced this is a bug in the processing.
Perhaps ACR is being spooked by the Leica's erratic exif metadata on aperture, and trying to compensate for an imbalance that is not there in the actual luminance or colour data in the files??? Perhaps in this specific instance that spooking is "confirmed" in the change of luminance values in the rocks between shots 1 & 2 and shot 3??? It does seem to want to transfer that tonal change down into the red dirt foreground.
This screenshot shows more clearly the two shots that ACR/LR is unable to blend, yet which PS can blend with ease:
This problem is very reminiscent of some of the issues that early iterations of the Photomerge command (PSCS2 iirc) used to throw up. I feel sure that it's a bug, and will in time get fixed by the ACR team. Just hope someone from Adobe reads this thread (are you there, Jeffrey T.?)
trshaner: I will try your suggestion and "Export to full-size ProPhoto RGB 16 bit TIFF", and I'll post back results later, but that just as an experiment, not as a potential workaround. The workaround is clear: use Photomerge in PSCC if LR can't do the job.
For now, though, I'm not prepared to give up on using in-camera Auto WB in the field, as to do so would mean I'd have to have an Expodisc with me at all times (pita). I think the ACR team can and will fix this in time, and while it persists in a small number of cases, I'll just round-trip to PS in the same old same old way. It's been a long time since Photomerge in PS let me down, and oftentimes I like to use Content-Aware Fill in preference to Autocrop anyway
Thanks for your input - very insightful, but I still think the problem's in ACR
Well, I tried the export as TIFF (16-bit, ProPhotoRGB uncompressed) as you suggested. Sadly with the same result...
I wonder if it could be the Leica's imaginative approach to writing aperture data into exif that's at the rotor the issue....???
(Making this up as I go along now)
Tried again, this time stripping out all metadata except copyright in the Export to TIFF procedure....
So it looks like the crazy Leica metadata can throw ACR off
Any comments welcome, but I think this could be the answer. Hopefully someone from Adobe will pick up on this. I'll put a link to this thread on the Leica forum
later unless a better explanation arises here
A small amount of WB difference and ~1EV exposure difference has never caused issues for me with Merge to HDR2 in PS and LR Merge to HDR should work the same. It does look some incompatibility with the Leica camera raw DNG files. If you can you post the the two camera DNGs with the issue (5875, 5876) to Dropbox or other download site I'll take a look at them.
For future HDR shots I still recommend using a fixed manual exposure and white balance setting across all images.
No HDR here, trshaner - just Merging to Pano
Thanks Jim! I was working on two different posts at the same time. I meant 'Merge to Panorama in PS' and LR 'Merge to Panorama.' The advice does apply to this post on Panorama errors.
I've put the files in my Dropbox and you should get a notice to the effect that you can download them
Look forward to your findings
Could you resend the share link via PM here. Thank you!
Got them. Give me a bit and I'll let you know what I find.
That's great Todd. Look forward to your findings, but take your time. No rush tips end!
A small amount of WB difference and ~1EV exposure difference has never caused issues for me with Merge to HDR2 in PS
Photoshop does seem to do a good job of blending exposure with varied WB and exposure. My nephew took a pano at Iguazu Falls several years ago and didn't set any thing on manual. The exposures and WB were all over the place and Photoshop still did a great job of blending. I also think it is still a good idea to set everything on manual though. Considering Lr Pano is fresh out of the box I think the Adobe team did a good job. Probably have some tweeks in the future versions.
The Leica M9 calculates the lens aperture F Stop value since the it is only mechanically connected to the camera:
Below is the DNG file EXIF data and LR's F Stop calculation.
I measured the two DNG images in "like pixel areas" near the right and left edges respectively. I used the LR WB tool with its largest area scale (17 x17 pixels) and the values are virtually identical exposures. However, LR's aperture calculation of F4.7 and F5.8 is a .85 EV difference.That means LR is applying .85 EV correction factor to the rightmost (M9-1005876.dng) image file prior to stitching and blending these two images. This makes the stitching function much more prone to control point errors and the blended image will have a .85 EV error at the stitch line.
LR Edit In> 'Merge to HDR Pro in PS' works fine with these images so clearly something is different with the way LR 'Merge to Panorama' handles the Leica M Series DNG files. I suggest you 'Report a problem' at the Photoshop Family forum here:
Place a link back to this post for reference.
Thanks Todd for confirming my hypothesis. Your demo of how LR handles the exif is most interesting to see, and I'm most grateful for your rigorous pursuit of this issue
"The Leica M9 calculates the lens aperture F Stop value since the it is only mechanically connected to the camera:" - This is as I said: the Leica M lenses have no electronic interface with the cameras, so these issues could presumably effect users of Leica M lenses on other mirrorless cameras too.
I will report to Adobe through the link you provide, and also post to the Leica forum, both with links back to here
So for the time being those of us who have PS as well as LR can still rely on PS to stitch where LR fails to do so, and hope Adobe can figure out a way of dealing with the issue
I'll check tomorrow how my Leica lenses work on my M43 cameras, and see if LR can stitch pans reliably in the absence of any metadata feedback on aperture settings, as they just show aperture readings of f/0.0 when you mount a Leica lens and (as they require) switch off lens recognition in the camera. If LR can deal with that, a bug-fix might just be to ignore the Leica M's' aperture data, which is purely conjectural anyway
Jim, please report back your findings. If that works (F0.0 EXIF) then it certainly is a workaround. A bug report is probably still a good idea and you can include your results with the fixed F0.0 EXIF Exposure values.
I've posted feedback to Photoshop.com here: LR/ACR Merge to Panorama has problem blending some files from Leica M9 camera
Will try some pan shots with Leica lens on M43 camera later today if it stops snowing
I've added my +1` vote and a comment.
I got a couple of gaps in the snowfall yesterday and ran a couple of tests.
Set both cameras the same: ISO 160, f/5.6, 1/250sec
First up a 4-image stitch from the M9 + 50mm Summicron:
Not easy to see the issue here in the small jpeg the forum allows, so here's a 100% crop from Library Loupe view of the trees top rightish:
I've tried several times, and it seems that the ACR/LR pano merge is much more consistent that PS ever is: repeated iterations were absolutely identical in every respect! In this instance the f/stop values of the 4 shots as recorded in the exif info were (l to r) f/6.7, f/8, f/5.7, f/9.5 - remember, in all cases the true aperture value was f/5.6!
Next a pano comprising 4 shots taken with my Pana GX7, with a Leica M-mount Voigtlander 25mm (= equivalent to 50mm on 35mm/ FF sensor). Camera's lens detection switched off, which results in camera recording aperture setting of f/0.0 in the exif info
No problems with the blending here, or in any other pano with this camera/ lens combination
Here's another from the M9 - much more successful, and comprising images with recorded f-stop values of (l to r) f/5.7, f/5.7, f/8, f/4.8.
While it is just about acceptable, it is clear that there's a tonal shift going on in the area directly above the top of the tallest tree, which is at the right-hand extreme of shot 2: see this screenshot of the 4 component shots in Grid view:
Best I got was this one, and it comes from blending 4 files with a pretty slight range of recorded f-stop values: (l to r) f/5.7, f/5.7. f/6.7. f/5.7 (ISO 160, 1/350sec)
if the others were as good as this, I'd have no problem. Here's one from the GX7 taken just seconds after the last (ISO160, 1/400sec, all f/0.0)
I rest my case: the Leica M9's eccentric approach to recording f-stops seems able to throw the new ACR/LR Merge to Pano's calculations awry. At least that's how it seems in my humble opinion and total ignorance of what goes on under the hood in digital photo processing apps
Maybe someone from the ACR Team might read this and make some sense of it - for now I've got a backlog of other work to do LR, and for now I can always round-trip to PS and do pan-stitching the old way. As I've said above I'm old enough to remember when Automate>Photomerge produced some bizarre, artefact-strewn results. But it got fixed and I'm sure this will too...
This LR plugins may be useful for setting the correct FNumber (Tag 0x9202) and ApertureValue (Tag 0x829d) in the EXIF data. I'm not sure which EXIF Tag LR uses for Merge to Panorama 'F stop' calculation, but worth a try.
I was thinking this evening that maybe exiftool could do this - never thought of lenstagger, which I have and use to correct lens identification when using non-Leica lenses on the M9.
But you're right: this, I think, and indeed it proves the point...
I'm not sure I understand what you're referring to here, Todd: "FNumber (Tag 0x9202) and ApertureValue (Tag 0x829d) in the EXIF data", but I set up Lenstagger like this (below), with f/5.6 in the F-Stop field and ran a couple of sets through it
Applying that to the original set of images of Kata Tjuta, I got this:
maybe not clear in that, so here is a blend of images 2 & 3 from the set (the locus of the original problem) now tonally mapped quite correctly:
as LR Merge to Pano is now reading all the files with aperture f/5.6!!!!
QED, I think
Many thanks Todd for all your input. Been fun figuring this one out!
By George I think you've nailed it! Until Adobe fixes this issue you've got a good workaround. Glad I could help.
You may have the same issue with Merge to HDR since it uses the exposure data for merging and blending the images. I would think LensTagger should work with LR Merge to HDR as well.
Haven't tried any HDR with the M9 yet. I don't use it much, but I think you're probably right, again!
I'd think all the ACR team need to do is put in some code to the effect of "if camera = Leica M digital, then aperture value = f/0.0" and the problem should just vanish overnight in ACR9.1/ LR6.1. Hopefully someone from that team may read this thread...
All the need to do is implement the same routine used in PS Photomerge for reading and applying the Exposure settings and it should work with out any EXIF editing.
You'd have a better understanding of these things than I, Todd
As to HDR, I've just tried a couple of these and it seems the problem may not arise here. I use the M9's menu settings for exposure bracketing, and it looks like the camera maybe behaves itself when that is switched on, and recognises that aperture will be fixed while shutter speeds are varied in accordance with the parameters (no. of exposures, shooting sequence and EV increments) set by the photographer. It would be possible to bracket "manually", I guess, but I can't imagine why anyone would want to do so! So a limited survey of my library suggests Merge to HDR may work OK
As long as LR is showing the same F/value for all of the bracketed image files then you should be OK. If not it may cause issues with the merged HDR DNG file during blending.
Agreed. So anyone using the camera's bracketing should be OK
I'm a member of the Camera Raw team. Are you willing to share a set of images that we can use to investigate this issue?
If so, please forward your email address to me at firstname.lastname@example.org and I will provide a location for you to upload the images.
I've sent you an invitation to my Dropbox where you'll find the files I used in the foregoing
I trust that will be ok
We'll take a look.
This issue seemed to have been resolved in ACR 9.1, but has resurfaced in more dramatic form in LR CC2015.2.1.
I started another thread here describing the problem, and it seems to effect not only the M9, but also the M Type 240, and therefore most likely all the Leica M digitals. It also effects not just Photo Merge> Panorama, but also Photo Merge> HDR, in both cases the processing seems to occur, but the preview and the output is just a solid black image, in the shape of the expected merged file output
Hopefully Chris you will see this thread: the files from last time are still available in my Dropbox, and I'll be happy to add some HDR files too, as well as output files from CC2015.2.1 if you get back to me